Intellectual Human-Object Classification Using Genetic Algorithm

Video surveillance applications need to segregate human from a variety of objects present in the environment to avoid the performance degradation in their classification. We propose a method to classify human and non-human objects present in the scene using Genetic Algorithm (GA). The initial population (chromosomes) for the genetic algorithm is formed by the coefficients of the Stationary wavelet transform applied for the image and evaluated using the proposed fitness function. Based on the best fitness values, the parents are selected for reproduction followed by mutation to repopulate the generations. The process of producing a new generation and selecting the best member is repeated until the best fitness valued chromosome is met. The Stationary wavelet transform (SWT) has multi resolution property that improves the reconstruction of the image when compared with the conventional Discrete Wavelet Transform (DWT). GA is a global stochastic search and optimization technique that can explore a large solution space and concentrate the search in the regions which lead to fitter structures and hence better solutions are obtained. The proposed method is evaluated by various performance metrics and found better than other state-of-art-methods.